LangGraph RAG & Search Agent with Ollama

This project demonstrates an AI agent built using LangChain, LangGraph, and Ollama. The agent can answer questions based on a provided resume (stored in data/cv.txt) using Retrieval-Augmented Generation (RAG) or search the web for external information using DuckDuckGo.

Agent Architecture

The following diagram illustrates the logical flow of the agent:

Features

• Conversational Agent: Interacts with the user via a command-line interface.
• Tool Use: Dynamically decides whether to use a tool based on the user's query.
• Resume RAG: Answers questions about resume content by embedding text chunks and retrieving relevant information.
• Web Search: Uses DuckDuckGo to find information about topics not covered in the resume.
• Ollama Integration: Leverages local Ollama models for both language generation and embeddings, ensuring privacy and offline capability.
• Modular Tools: Tool logic is separated into a tools directory.

Prerequisites

1. Conda: Ensure you have Conda installed. This is used for managing the Python environment. You can install it from Anaconda Distribution or Miniconda.
2. Ollama: Install Ollama by following the instructions on the Ollama website.

Setup

1. Clone the Repository (Optional): If you haven't already, clone the repository:

   git clone <your-repo-url>
   cd <repository-directory>

2. Install and Run Ollama Models: Make sure the Ollama application/server is running. Then, pull the required models (LLM and embedding model):

   ollama pull llama3.2 # The LLM used by the agent (can be changed in agent.py)
   ollama pull nomic-embed-text # The embedding model used for RAG

   Important Note: Ollama must be running in the background before starting the agent.

3. Create and Activate Conda Environment: Create a new conda environment (e.g., named agent-env) with a specific Python version (tested with 3.12):

   conda create -n agent-env python=3.12 -y
   conda activate agent-env

4. Install Python Dependencies: Install the required Python packages using pip and the requirements.txt file:

   pip install -r requirements.txt

5. Prepare Resume File: Place your resume content in a file named cv.txt in the data directory of the project. A sample file is included, but you should replace it with your actual resume text for the RAG tool to work correctly.

Running the Agent

1. Ensure Ollama is running.
2. Activate the Conda environment:

   conda activate agent-env

3. Run the agent script:

   python agent.py

How to Use

Once the agent starts, it will prompt you for input. You can ask questions like:

• "What was my most recent job?" (Should trigger the resume_rag_tool)
• "Tell me about LifeMine Therapeutics." (Should trigger the web_search_tool)
• "What is LangGraph?" (Should trigger the web_search_tool or be answered directly by the LLM)

Type quit or exit to end the conversation.

Project Structure

.
├── agent.py            # Main agent script with LangGraph logic
├── data/               # Directory for data files
│   └── cv.txt          # Resume file (replace with your own)
├── docs/               # Documentation assets
│   └── agent_graph.png # Agent architecture diagram
├── requirements.txt    # Python dependencies
├── tools/              # Directory for tool implementations
│   ├── __init__.py
│   ├── resume_rag.py   # RAG tool implementation
│   └── web_search.py   # Web search tool implementation
└── README.md           # This file